ResearchEngineering/financial_analyst

📊 Portfolio Volatility Analyzer

Analyze your investment portfolio risk using Modern Portfolio Theory with OCR, interactive visualizations, and beautiful mathematical formulas.

Features

• —📸 OCR Portfolio Parsing: Upload screenshots of your portfolio and automatically extract tickers and amounts
• —✏️ Editable JSON: Correct OCR errors with an intuitive JSON editor
• —📈 Historical Data: Automatically fetch 1 year of price data from Yahoo Finance
• —🧮 Full Calculations:
• —Portfolio weights
• —Log returns
• —Covariance matrix
• —Portfolio variance and volatility
• —📐 Beautiful LaTeX Formulas: See every calculation step with symbolic and numerical formulas
• —📊 Detailed Variance Expansion: Step-by-step breakdown showing how each asset contributes to portfolio risk
• —🎚️ Interactive Rebalancing: Adjust portfolio amounts with sliders and see volatility update in real-time

How to Use

1. 1.Upload Portfolio Screenshot: Take a screenshot of your portfolio (must show ticker symbols and dollar amounts)
2. 2.Edit Portfolio JSON: Review and correct any OCR errors in the JSON editor
3. 3.Validate Portfolio: Click "Validate Portfolio" to start analysis
4. 4.View Results: See historical data, covariance matrix, and detailed formulas
5. 5.Rebalance: Use interactive sliders to adjust positions and see impact on volatility

Technical Details

Formula Highlights

Portfolio Variance:

σ²_p = w^T × Σ × w

Where:

• —w = vector of portfolio weights
• —Σ = covariance matrix (annualized)

Portfolio Volatility:

σ_p = √(σ²_p)

Architecture

• —Frontend: Streamlit
• —OCR: Tesseract (pytesseract)
• —Financial Data: yfinance (Yahoo Finance)
• —Math: NumPy, Pandas, SymPy
• —Deployment: Docker on Hugging Face Spaces

Local Development

Prerequisites

• —Python 3.11+
• —Tesseract OCR installed

Setup

# Install dependencies
pip install -r requirements.txt

# Run the app
streamlit run app.py

Docker Build

# Build
docker build -t portfolio-analyzer .

# Run
docker run -p 7860:7860 portfolio-analyzer

Example Portfolio

Test the app with this JSON:

{
  "AAPL": 5000,
  "GOOGL": 3000,
  "MSFT": 2000
}

Notes

• —Uses 252 trading days for annualization
• —Calculates log returns: ln(Pt / P{t-1})
• —Smart truncation for portfolios with 4+ tickers
• —1-hour cache for historical data to reduce API calls

Built With

• —Modern Portfolio Theory
• —LaTeX mathematical notation
• —Real-time financial data

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference